Trait‐based life strategies, ecological niches, and niche overlap in the nekton of the data‐poor Mediterranean Sea

Abstract Biological traits can determine species ecological niches and define species responses to environmental variation. Species have a specific functional position in the biological community, resulting in interactions like interspecific competition. In this study, we used biological traits in order to define the life strategies of 205 nektonic species of the Mediterranean Sea. Furthermore, traits related to resource use were analyzed to determine the level of trait and niche overlap and their relationship to life strategies. Focusing on habitats of importance (Posidonia beds, coralligène formations, and lagoons), we investigated strategies and niches of the species present there. Finally, we examined the life strategy of Lessepsian species and investigated the niche overlap between them and indigenous species. Archetypal analysis indicated the existence of three life histories corresponding to strategies already documented for fish (equilibrium, periodic, and opportunistic), with some species also placed in intermediate positions. Niche overlap was evaluated by multiple correspondence analysis and the generation of a single distance metric between all species pairs. This identified species occupying relatively empty (underexploited) ecological niches, like the Lessepsian species Siganus luridus and S. rivulatus, a finding that can also be associated with their establishment in the Mediterranean. Most Lessepsian species were associated with the opportunistic life history strategy, again an important aspect related to their establishment. Also, we documented that most species occurring in important habitats have a relatively high overlap of niches. No significant differences were found in the life strategies across Mediterranean habitats; however, variation in niche overlap and traits related to habitat use was detected. The findings can be useful to determine theoretical competition between species and to identify empty ecological niches. Fisheries science can also benefit from comprehending the dynamics of competing stocks or predict the responses of data‐poor stocks to anthropogenic stressors from known examples of species with shared life strategies.


| INTRODUC TI ON
Organisms possess various traits which relate to their ability to survive, grow, and produce offspring, that is, to increase their fitness (Violle et al., 2007). The term trait refers to the various aspects of the biology of an organism (e.g., physiology and behavior) that characterize its population responses to environmental changes or its role in ecosystem processes (Diaz & Cabido, 2001;Violle et al., 2007).
The combinations of the various biological traits of a species form its life history strategy. Pianka (1970) has identified two main life strategies: The r-strategy incorporates elements like fast growth, high fecundity, and mortality, while the K-strategy conversely combines slow growth, low fecundity, and lower mortality rates.
Regarding fish, Winemiller and Rose (1992) have indicated three strategies: the equilibrium, periodic, and opportunistic strategy.
The life history strategies mentioned above allow fish and nektonic populations to exploit the resources available in their environment in the best possible way (Pianka, 2006); however, the limitation of resources in the ecosystem results in competition between different species. In fact, when the community is in balance, the combination of a species' functional characteristics can describe its role in the community, thus its ecological niche (Cadotte, Arnillas, Livingstone, & Yasui, 2015;Violle & Jiang, 2009). In cases of niche overlap between different species, there is competition for the use of common resources (Hutchinson, 1957;Pianka, 2006).
As different species may sometimes serve the same functions (sometimes at similar or different rates- Duffy, Shackelton, & Holmes, 2008), an approach where species' functional roles are examined using traits is of particular importance, especially in areas characterized by high biodiversity. The Mediterranean Sea is considered a biodiversity hotspot (Myers, Mittermeier, Mittermeier, da Fonseca, & Kent, 2000) as, despite its surface being lower than 0.7% of the global ocean, 4%-18% of global marine species are estimated to be distributed there (Bianchi & Morri, 2000). Furthermore, the Mediterranean is rapidly colonized by Lessepsian species (moving from the Red Sea to the Mediterranean across the Suez Canal- Por, 1978); thus, biodiversity patterns are rapidly changing. Arndt and Schembri (2015) have indicated that specific traits (like size, spawning type, and tendency to form schools) are related to the establishment and spread of Lessepsian fish species in the Mediterranean. The Mediterranean hosts a variety of habitats, some of which are characterized by a very high biodiversity, like Posidonia oceanica beds and coralligène formations. These habitats are widely distributed in the Mediterranean basin and have an important role in ecological processes providing shelter or being nursery areas (e.g., Ballesteros, 2006;Kalogirou, Corsini-Foka, Sioulas, Wennhage, & Pihl, 2010). Lagoons are another important habitat type as, being eutrophic in the generally oligotrophic Mediterranean, they are highly productive, and act as fish nursery grounds (Viaroli et al., 1996).
The examination of the biological traits of the species distributed in these habitats could provide information about their role and the interspecific relationships existing there which in turn may be useful for conservation efforts.

Despite the increasing number of publications about
Mediterranean nekton (mainly fish), and the anthropogenic effects on nektonic populations, many aspects of the marine ecosystem are yet unknown and attempts to apply the so-called Ecosystem Approach (Garcia, 2003) are relatively recent and exploratory (e.g., Link, Huse, Gaichas, & Marshak, 2020). In this data-poor region, the definition of ecological niches and the determination of interspecific relationships are important challenges (Givan, Parravicini, Kulbicki, & Belmaker, 2017). While community assembly and interspecific relationships are generally investigated using abundance data of the species present in biological communities (e.g., Pecuchet, Törnroos, & Lindegren, 2016), an approach using bibliographic data on species traits could help indicate relationships that are otherwise neglected (because of the species' relative abundance) or highlight potential interactions that are hidden due to the particularities of the ecosystem studied.
The aim of this study is to use biological traits in order to

| Species selection
We used a subset of a broader dataset of 23 traits of 235 species present in the Mediterranean nekton. Even though the inventory of nektonic species of the Mediterranean Sea is larger, we attempted to include as many species as possible to have a representative depiction of the community, covering pelagic, benthic and benthopelagic species (Tzanatos, Moukas, & Koutsidi, 2020). Special attention was given to include all species found in two Mediterranean habitats known for the biodiversity they host, specifically 66 species found in Posidonia beds (Fernandez, Milazzo, Badalamenti, & D'Anna, 2005;Francour, 1997;Guidetti, 2000;Kalogirou et al., 2010;Moranta, Palmer, Morey, Ruiz, & Morales-Nin, 2006) and 26 found in coralligène (Ballesteros, 2006). We also included 28 species found in lagoons, as they are highly productive ecosystems (Nicolaidou, Reizopoulou, Koutsoubas, Orfanidis, & Kevrekidis, 2005

| Life strategies
For the identification of ecological niches related to the life history of the species examined, we used Archetypal analysis (Cutler & Breiman, 1994) in a process similar to that followed by Pecuchet et al. (2017). The Archetypal analysis represents each individual in a dataset as a mixture of "individuals of pure type" or "archetypes." These are a small number of (not necessarily observed) extreme points in a set of multivariate observations; the data are expressed as a probabilistic mixture of archetypes (Cutler & Breiman, 1994;Eugster & Leisch, 2009;Li, Wang, Louviere, & Carson, 2003). As this analysis uses continuous data, five continuous traits of the dataset relevant to life history definition were used: longevity, age at maturity, fecundity, maximum length, and trophic level. Longevity and fecundity were log-transformed. All five traits were scaled, to ensure equal weights.
As there is no rule for the correct number of archetypes k, we determined its value by running the algorithm for k = 1, 2, … 10, by performing 10 iterations and calculating the residual sum of squares (RSS) for each one (Eugster & Leisch, 2009). A small value of k with low RSS was chosen, according to the "elbow criterion." The R library "archetypes" was used for this analysis (Eugster & Leisch, 2009).

| Shared traits, ecological niches, and niche overlap
The identification of species' ecological niches is important in order to determine the role of each species in the environment, and to detect possible competition in case of niche overlap (Pianka, 2006).
The concept of competition refers to interaction regarding the exploitation of specific resources, and the traits examined in life strategies are not necessarily relevant to resource use. For this, we used 10 traits relevant to resource exploitation. These traits were divided in three groups based on the type of the main resource use with which these traits can be linked: (a) Maximum length, trophic level, diet and feeding type were used to describe the use of food resources by mature fish, (b) optimal depth, depth range, habitat type, and seabed type were used as a descriptor of habitat use, and (c) spawning period and spawning habitat were considered to reflect spawning habitat use (even though behavioral aspects of some species, for example, carrying their eggs may render the interactions more complex). The two traits: spawning period and spawning habitat can also be relevant for food resource use by larvae and juveniles (as, e.g., nektonic larvae coinciding spatially and temporally can be expected to use the same food resources). Separate species groupings based on sharing the traits relevant to each resource indicated the overlap of simple niches described by the relevant traits. It also indicated trait combinations that were devoid of coverage by any species.
In order to create a more synthetic image and determine traitsbased ecological niches following the "hypervolume" approach (e.g., Pianka, 2006 Pianka, 2015). The traits' contribution to each dimension is presented in the Figure S1.
To estimate the niche overlap between any two species, their coordinates in the four MCA dimensions retained were taken into account. For this, the distances between all possible pairs of 205 species for each of the four retained MCA dimensions were calculated and standardized resulting in four triangular distance matrices. The relevant contribution of each dimension to % variance explained was used to weight the distances of the four triangular matrices (each corresponding to a dimension retained) producing a single weighted distance metric between all species pairs in a final triangular matrix. The smaller this single distance was between two species the higher niche overlap they were considered to have for all resource types. Pairs of species with high distance values were on the contrary considered to have low niche overlap.
To determine whether there are significant differences in the average distances/level of niche overlaps between the different life strategies previously identified, a PERMANOVA (Permutational multivariate analysis of variance) was carried out. As a next step, to determine which groups significantly differed from others, an ANOSIM (Analysis of similarities) on the triangular distance matrix of the MCA dimensions by species was performed. While PERMANOVA can also be designed to indicate pairwise contrasts, ANOSIM was also included, as its R statistic is an absolute measure of the strength of the between-groups difference, contrary to the PERMANOVA pseudo-F (Anderson, Gorley, & Clarke, 2008).
Whether niche overlap is high or low must first be measured relative to some null expectation. For this, the library of R, "EcoSimR" (Gotelli, Hart, Ellison, & Hart, 2015) was used to test the overlap in resource use among the set of 205 species (36 modalities of the 10 traits used). The analysis reveals whether the average niche overlap, calculated among all unique pairs of species, is more or less than would be expected if species used resource categories independently of one another. Here, the "pianka" niche overlap index was used, which ranges from 0 (no overlap) to 1 (complete overlap) indicating the mean overlap of all possible species pairs.

| Comparisons between species present in important Mediterranean habitats
The archetypal analysis identifies archetypes and provides the percentage that each species is characterized by each archetype.
In order to detect whether there are differences between the species assemblages present in important Mediterranean habitats, and possible association(s) with a specific archetype, we compared the arcsine-transformed percentage scored for each of the three archetypes identified between the groups of species distributed in Posidonia beds, coralligène formations, and lagoons, using the Kruskal-Wallis test, as the parametric prerequisites were not met. In the case that a species appeared in more than one habitat, its score was repeated for all habitats in which it was found.
Furthermore, a PERMANOVA was carried out to detect whether there are differences in the average distances/level of niche overlaps of the species present in Posidonia beds, coralligène formations, and lagoons. Again, as a next step, to determine which groups significantly differed from others, an ANOSIM on the triangular distance matrix of the MCA dimensions by species was performed. Also, to highlight the differences in niche overlap between the species assemblages in Posidonia beds, coralligène formations, and lagoons, a SIMPER (Similarity percentage analysis) on the triangular distance matrix of the MCA dimensions by species was performed.

| Comparisons between indigenous and Lessepsian species
In order to detect whether Lessepsian species tend to be associated with a specific archetype, we compared the arcsine-transformed Species assigned to an archetype should have a percentage higher than 50% for this archetype, and the percent difference between the first and the second archetypes should be higher than 15%.
Species intermediate between two archetypes should have a percent difference between the two highest archetypes lower than 15% and the percentage of the third one lower than 20%. Finally, to assign a species as intermediate between all archetypes the minimum percentage of any archetype should be higher than 20% and the maximum lower than 50%.
As a next step, to determine which groups significantly differed from others, an ANOSIM on the triangular distance matrix of the MCA dimensions by species was performed.
All inferential tests were carried out at the α = 0.05 significance level.

| Life strategies
Using the "elbow criterion" (Figure S2), the optimal number of archetypes (niches associated with species life history patterns) was three ( Figure 2). An alternative would be to determine five archetypes; however, according to the law of parsimony (also known as "Occam's razor") that renders more possible the existence of a lower rather than higher number of archetypes, we proceeded with three archetypes.
The proportional affinity of each species to the three archetypes identified (Table S1 and Figure S3) Figure S4). Niche overlap ranged from 0% to 74% (the latter being the highest distance found between any two species, note that the range is standardized from 0% to 100% in the Interactive Figure).  (Figure 3a) is also confirmed by their relatively high average distance in relation to its standard deviation (Figure 4).
The PERMANOVA indicated the existence of significant differences in average MCA distances (i.e., level of niche overlaps) between strategies (pseudo-F = 12.573, p = .001). The ANOSIM also indicated significant differences between pairwise strategies. Species belonging to archetype-1 were significantly different in trait composition from all the other strategy levels (

| Life strategies and ecological niches of species present in important habitats
The archetype affinity scores of species distributed in the three important Mediterranean habitats (Posidonia beds, coralligène, and lagoons) were not found to significantly differ regarding either the first The PERMANOVA indicated significant differences in the average MCA-weighted distances (niche overlap) of the species distributed between habitats (pseudo-F = 3.8723, p = .008). The ANOSIM also indicated significant differences in the MCA dimensions of the species between pairs of habitats: Species appearing in lagoons were significantly different in trait composition from those of coralligène assemblages (R statistic = 0.053, p = .042), albeit the relationship was weak. SIMPER analysis indicated an average dissimilarity 58.43% between lagoons and coralligène formations. The modalities responsible for the dissimilarity for the species distributed in lagoons were as follows: trophic level: 3, optimal depth: 0-50 m, sea bed type: soft, spawning habitat: pelagic, and spawning period: summer and spring, while the modalities that were responsible for the dissimilarity for the species over coralligène formations were as follows: trophic level: 4, all ranges of optimal depth, sea bed type: variable, spawning habitat: pelagic and benthic, and spawning period: summer.
Examining the niche overlaps between species within the as-

| Life strategies and ecological niches of Lessepsian species
Regarding life strategies, the 22 Lessepsian species were found to Lessepsian species can be clearly associated with the first archetype (also indicated in Figure 2 and in Table S1).
Concerning the Lessepsian species examined, most of them were within a distance ranging from 0% to 40% with most (>70%) of the indigenous ones (Figures 4 and 5d). However, there were three exceptions indicating low niche overlap with indigenous species: Scomberomorus commerson had an average distance of 0%-40% with half of the indigenous species and with the other half a distance higher than 40%. Siganus luridus and Siganus rivulatus had a distance higher than 50% with at least 70% of the indigenous species. Overall, however, comparing the average niche overlaps of the Lessepsian species against those of the indigenous ones did not indicate significant differences between these two origin groups (pseudo-F = 1.0174, p = .355).

| D ISCUSS I ON
In the present study, we analyze five life cycle biological traits to detect the life history strategies of 205 Mediterranean nektonic species. We document three life history strategies. Using 10 biological traits related to resource exploitation, we define ecological niches and detect niche and trait overlap (both overall and concerning specific resource types), generating a unique distance metric to assess niche overlap between all species pairs. Taking a closer look at Lessepsian species, we document their affinity to the first life strategy and examine niche overlap between them and indigenous species indicating, for example, the relatively empty ecological niche occupied by the herbivorous invaders Siganus luridus and S. rivulatus.
Furthermore, we document that most species occurring in important habitats have a relatively high niche overlap with species occurring in the same habitat ( Figure 5).
As stated above, we identify three main life history niches of Mediterranean nekton. Even though in this study, we do not use the same traits as the theoretical framework for the detection of life history strategies (Pecuchet et al., 2017;Winemiller & Rose, 1992)-as there is absence of the traits "parental care" and "offspring size" in our dataset-the species characteristics occupying the three detected life history niches documented here (red, green, and blue in Figure 2) are similar to the characteristics of the species following the life history strategies defined there: opportunistic, equilibrium, and periodic. The species following the first life history niche correspond to the opportunistic strategy. They have a short generation time, thus show a high intrinsic growth rate, despite their relatively low individual fecundity (King & McFarlane, 2003). Additionally, they have high natural mortality (Beverton & Holt, 1959). The group of species in this strategy comprises mainly small pelagic or coastal, herbivorous benthopelagic species associated with habitats defined by disturbance and high variability, but also with high energy resources (King & McFarlane, 2003). Most Lessepsian species are associated with this strategy in our work.
The species of the second life history niche are similar to the equilibrium strategy of Winemiller and Rose (1992). There are mostly chondrichthyans, are characterized by late maturation (extended brooding season increases the risk of mortality before reproduction) and low fecundity (low dispersal), traits that make them vulnerable to environmental changes (Stevens, Bonfil, Dulvy, & Walker, 2000) and human impacts (Lucifora, García, Menni, & Worm, 2012). Thus, they are favored in stable habitats, with low environmental changes (Mims & Olden, 2012). They have also been documented to show low resilience to fishing mortality and low recovery rates after population decline (Hoenig & Gruber, 1990;Stevens et al., 2000).
Most chondrichthyans are widespread marine food chain top predators. Because of their rare traits combinations (the rarity of low fecundity has been documented by Koutsidi, Tzanatos, Machias, &Vassilopoulou, 2016 andTzanatos et al., 2020), their removal could result in species replacement and niche vacancy, potentially influencing ecosystem structure and functioning (Gouraguine et al., 2011).
The third life history niche is similar to the periodic strategy (Pecuchet et al., 2017;Winemiller & Rose, 1992) and included large bodied species with long life span and high fecundity. Species with these traits (e.g., Polyprion americanus and Pseudocaranx dentex) are able to cope with variable and unpredictable environments by producing large numbers of eggs (Hutchings, 2002). Thus, this niche includes species likely to be favored in highly seasonal environments (Mims & Olden, 2012). High longevity and fecundity enhance these species' survivorship though low productivity regimes and storage energy for future more favorable environments (King, McFarlane, & Beamish, 2001). It is important to note that according to our findings (Table S1) Regarding the overlap in both trait groups related to a specific resource independently (adult food, adult habitat, and spawning habitat) and in the ecological niches described by combining all 10 traits, it is important to note that some combinations of traits/niches may be empty because no species with these trait combinations is distributed in the study area (or was not included in the dataset analyzed).
However, there is also the possibility that the specific traits set does not exist, as it is known that traits can exist in specific combinations following the main life strategies (e.g., King & McFarlane, 2003 and also documented here); thus, for example, combinations of small size with high trophic level are not found because they are not possible, not because the actual niche is vacant.
Many species have relatively small distance from others, and thus, there is relatively high niche overlap between these two species. The examination of the distances, not on a pairwise level, but for the entire assemblage, can be an indication of community assembly rules through competitive interactions. This interpopulation process may eventually result in combinations of species that will continue to coexist (McGill, Enquist, Weiher, & Westoby, 2006).
However, further insights on the community assembly rules would need to use actual abundance (and not presence only) data as this would allow the quantification of interspecific and intraspecific competition. Furthermore, it should be noted that here we evaluate niche overlap. This overlap may be linked to potential competition based on the availability of the relevant resource (e.g., food for adult stages). However, for actual interspecific competition to exist, the different species need to co-occur in the same habitat.
Even if two species co-occur and group together (regarding the use of a specific resource), this does not necessarily reflect competition.
Resource abundance has also to be taken into account since these species could be exploiting a very common, plentiful food resource (Brocksen, Davis, & Warren, 1968).
Regarding specific habitats (Posidonia beds, coralligène, and lagoons), more species with a high niche overlap with others were found to exist in Posidonia beds (but the number of species occurring there is anyway higher). Species distributed in these important Mediterranean habitats have no significant affinity with a specific life history strategy, as generally the opportunistic and periodic strategies are common and the equilibrium strategy is rare in all three habitats while "intermediate" species are also generally common. This is an interesting finding and may indicate that there is no environmental filtering tending to select only one strategy to prevail over a specific habitat, but the communities are composed of species whose fitness can be attained by a combination of different adaptations, that is, following different strategies (McCann, 1998 (Ballesteros, 2006;Nicolaidou et al., 2005), thus indicating some environmental filtering on traits related to habitat properties.
Concerning Lessepsian species, the lack of significant difference between them and indigenous ones in their average distance from all other species is complemented by the finding that most have high niche overlap with many native species (PERMANOVA results and Figure 5d). Arndt, Givan, Edelist, Sonin, and Belmaker (2018) (Figures 4 and 5d) probably explains their establishment due to their ability to occupy underexploited or vacant niches (Mavruk & Avsar, 2008;Por, 1978).
As mentioned above, in the present work most Lessepsian species are related to the first life strategy. This possibly makes these species capable of adapting to environments characterized by variability (King & McFarlane, 2003) and could be key for their spread and establishment in the Mediterranean, as the ability to adapt to various environmental conditions is generally accepted as a precondition for successful establishment of invaders in marine habitats (Moyle & Marchetti, 2006;Safriel & Ritte, 1980 combinations or rare modalities may be keystone species, that is, species whose relative importance for the community is much greater than their relative abundance (Bond, 2001). The decline, or in the extreme case, extirpation of their populations, may deprive the ecosystem from ecological functions and lead to the creation of vacant niches. In general, the existence of vacant niches (nonshaded areas in Figure 3b, but also see Figure 4, Figure S4) in the assemblage could potentially increase the possibility of the spread and establishment of invading populations (Belmaker, Parravicini, & Kulbicki, 2014;Givan et al., 2017 widespread discussion about the optimal number of traits to use (Pecuchet et al., 2017); thus, perhaps our findings would be different if we used a different number of traits. This is also the reason for selecting only traits that can be clearly associated with resource use and also distinguishing different resource uses in analyses related to species roles and competition.
Fishing gears can be associated with specific traits, selecting species that possess them and thus decreasing the trait presence in the community (Koutsidi et al., 2016;Mbaru, Graham, McClanahan, & Cinner, 2020). Naturally, the exploitation of stocks through fisheries has direct (by removing the catches of some species) and indirect (by changing predation and competition patterns due to the relative change in the abundance of different populations in the community) impacts on the community composition. This may in turn have significant implications for the traits composition and the associated functioning at the community level and could be examined in future modeling simulations.
The findings of the present work could extend to study community assembly rules using objective abundance data (e.g., survey data) in the future. Additionally, the niche overlap documented here could help establish an understanding of species population dynamics and be used in fisheries management (King & McFarlane, 2003), for example, by incorporating into fisheries models the dynamics of species competing with the stock under examination (as, e.g., sometimes the "gap" left by the reduction in the abundance of target species can be filled by competing unex- for comments on preliminary results. We would also like to thank two anonymous referees for their comments that helped improve the manuscript.

CO N FLI C T O F I NTE R E S T
The authors declare no conflict of interest.

DATA AVA I L A B I L I T Y S TAT E M E N T
Data are available from figshare (Koutsidi, Moukas, & Tzanatos, 2018